Intelligent jamming decision-making system based on reinforcement learning

Abstract

Cognitive communication countermeasures have increasingly been emphasized as an important research interests in cognitive electronic warfare. However, the low signal-to-noise ratio (SNR) and frequency hopping (FH) in communication countermeasures create significant difficulties for spectrum sensing and jamming decision-making. In this paper, an intelligent jamming decision-making system for FH communication is designed based on an improved deep Q- network (DQN). First, a spectrum sensing method utilizing a bidirectional long short-term memory (Bi-LSTM) network is introduced, which establishes the received signals as a binary hypothesis testing model and employs the Bi-LSTM network for signal classification. Second, the jamming channel selection problem is modeled as a Markov decision process (MDP), and an improved DQN algorithm is applied to facilitate intelligent decision-making for jamming channels. Finally, simulation experiments are conducted to evaluate the performance of the algorithms. The results show that the proposed Bi-LSTM network achieves a detection probability of over 88% even in low-SNR communication countermeasure environments at $-12$ dB. Furthermore, the improved DQN algorithm achieves a 100% channel jamming rate and the fastest convergence speed among the five compared algorithms, effectively learning the FH sequences and implementing jamming.